ELINOR NEMLANDER - Development of risk assessment tools to aid in early detection of cancer in primary health care

Project description

Most cancer investigations start in primary health care (PHC) based on symptoms and signs making the patient, the PHC physician, or both to suspect an underlying malignancy. Evidence-based methods for assessing patients’ risk for cancer prior to the investigation are lacking in clinical practice today.

We intend to investigate the predictive value of symptoms for cancer in PHC to find ways to improve early cancer detection. The overall aim is to create a basis for the development of evidence-based risk assessment tools for cancer. We will do this by:

Investigating the ability of anaemia as a marker for early identification of patients in primary care with cancer. We will build a cohort of individuals in the Stockholm region, who underwent haemoglobin assessments during study period and acquire data from three different data sources. Models for the ability of anaemia to discriminate cancer will be constructed using a machine learning model and cox regression analyses.

Investigating how general practitioners can identify patients with lung cancer based on symptoms, clinical findings, and diagnostic profile. We will analyze questionnaires from patients with suspected lung cancer from Karolinska Hospital. Machine learning analyzes will be performed on data to find possible combinations of symptoms and findings that can predict lung cancer in never-smokers. The method will then be tested nationally on never-smoking patients who have recently been diagnosed with lung cancer and matched controls.

Methods or tools for assessing a patient’s cancer risk in PHC based on their symptoms can contribute to earlier cancer detection. If we can limit investigations to patients most at risk, this may also reduce unnecessary diagnostic procedures and the risk of displacement.